Linearity: The Relationship Must Be Linear
You'll learn why simple linear regression only works when X and Y relate in a straight-line pattern.
The Core Idea
Simple linear regression assumes that the relationship between your predictor variable (X) and your outcome variable (Y) can be represented by a straight line. This is the linearity assumption—arguably the most fundamental requirement of the model.
If the true relationship between X and Y curves, bends, or follows any non-straight pattern, forcing a straight line through it will produce poor predictions and misleading results. Your residuals will show systematic patterns rather than random scatter.
Why This Matters
Think of it like navigating a winding mountain road with only straight-line directions. You might get the general direction right, but you'll constantly be off-course at the curves. Similarly, if you fit a linear model to a curved relationship:
- Your predictions will be systematically wrong in predictable ways
- Your R-squared will underestimate the true strength of the relationship
- Your slope and intercept estimates will be misleading
- Confidence intervals and hypothesis tests become unreliable
Checking for Linearity
The best initial check is visual: create a scatterplot of X versus Y before fitting your model. Look for a pattern that roughly follows a straight line. After fitting, examine the residual plot (residuals vs fitted values)—if you see a curved pattern or funnel shape, the linearity assumption is violated.
Key Takeaway: Simple linear regression requires that the relationship between X and Y follows a straight line; if the true pattern is curved or non-linear, the model will fail to capture the relationship accurately and produce unreliable results.